In the conventional control, the feedback control of PID control has been generally used. In the PID control, a control output is always determined with a delay from a phenomenon, and therefore, if each control gain of PID is increased in an attempt to increase the control speed, the control no longer catches up with the phenomenon, and therefore, the control becomes unstable. In particular, if a mechanical damping force of a controlled object reduces remarkably, the control tends to become unstable and there is a case where the control diverges. As a method for determining each control gain of the PID control in order to avoid the instability of control, a control theory, such as an H∞, capable of ensuring the stability of control is applied. However, under the restrictions of the PID control, overshoot or control delay occurs due to load fluctuation.
In the PID control also, if sliding mode control is used, it is possible to theoretically eliminate the influence of load fluctuation by switching control gains in accordance with the control state. However, if the control period is lengthened, this control keeps oscillating and no longer converges. Because of this, in order to completely eliminate the influence of load fluctuation, it is necessary to switch control gains at infinitely high speed, and control at speed that can be said as infinitely high speed for a phenomenon will be necessary. Further, adjustment of each control gain, such as PID, is necessary and the quality of the adjustment of control gain determines the quality of control, and therefore, the adjustment of control gain becomes a very important factor.
Furthermore, these control theories are for making up for faults of the PID control and are not methods designed for the purpose of control to “cause a controlled object to stay at a target position in the shortest time”. Because of, this, for this simple purpose, it can be said that the time optimal control is a control method more suitable to the purpose rather than the PID control.
The simplest time optimal control is a control to cause a controlled object to stay at a target position by accelerating the controlled object by a maximum thrust force up to the middle on the way to the target position and by decelerating it at a maximum deceleration for the rest of the way. This output pattern is determined before the control is started, and therefore, the time optimal control can be referred to as feedforward control.
In other words, the time optimal control is a control method for moving a controlled object by a maximum driving force of an actuator and stopping it by a maximum braking force, and is theoretically a control capable of causing the controlled object to stay at the target in the shortest time. That is, the time optimal control is a control method that perfectly meets the purpose of control to “cause a controlled object to stay at a target position in the shortest time”.
For example, as described in Japanese patent application Kokai publication No. 2000-94371, as a control device using the time optimal control, the time optimal control device of a robot is proposed, which includes a control unit configured to control a servomotor, a correspondence relationship storage unit configured to store a relationship between a controlled variable on the basis of the value at the time of no load and a load weight, a load estimation calculation unit, an acceleration/deceleration constant determination unit configured to determine acceleration/deceleration constants based on workpiece information calculated by the load estimation calculation unit, and a command creation unit configured to create a command to be delivered to the servo control unit using the determined acceleration/deceleration constants, and which lengthens the acceleration time when grasping a workpiece and shortens the acceleration time when not grasping a workpiece.
However, while the time optimal control is theoretically an ideal control capable of control with the shortest time, it is an open control in which the output pattern is determined by taking into consideration the initial velocity, the maximum acceleration, and the maximum deceleration, and because there is no feedback element, there is such a problem that no modification method is available when the target and the controlled variable do not agree and it is difficult to cause the target and the controlled variable to agree accurately, and therefore, it is rarely adopted in actual control.